CN114529755A

CN114529755A - Tool and system for testing papillary thyroid carcinoma

Info

Publication number: CN114529755A
Application number: CN202210082476.9A
Authority: CN
Inventors: 李铮; 任夏萌; 舒健; 于汉杰
Original assignee: Northwest University
Current assignee: Northwest University
Priority date: 2022-01-24
Filing date: 2022-01-24
Publication date: 2022-05-24
Anticipated expiration: 2042-01-24
Also published as: CN114529755B

Abstract

The invention provides a tool and a system for testing papillary thyroid carcinoma, which comprise: a processor and a storage medium; the storage medium performs data interaction with the processor, and is used for executing the following steps when the program stored in the storage medium is loaded by the processor: and identifying the lectin chip data of saliva of the patient to be diagnosed by a papillary thyroid carcinoma identification model, and determining whether the patient to be diagnosed is the papillary thyroid carcinoma patient. The tool for testing papillary thyroid carcinoma disclosed by the invention is used for identifying saliva of a patient to be diagnosed through the papillary thyroid carcinoma diagnosis module and determining whether the sample to be diagnosed is a papillary thyroid carcinoma sample, has the characteristics of convenience in sampling and high sensitivity, and can be used for quickly identifying whether a subject suffers from papillary thyroid carcinoma.

Description

Tool and system for testing papillary thyroid carcinoma

Technical Field

The invention relates to the field of biological computers, in particular to a tool and a system for testing papillary thyroid carcinoma.

Background

Thyroid Cancer (TC) is the most common malignant tumor of the endocrine system, and the incidence rate thereof is rapidly increased with the progress of diagnostic technology, and according to the global cancer statistical report in 2020, thyroid cancer ranks ninth in the incidence rate of the first ten cancers with over 58 ten thousand new cases. Papillary carcinoma (PTC) is the most common subtype, with 80% of new cases being papillary carcinoma with the best overall prognosis, with metastases most common in the cervical lymph nodes and less distal involvement. Early monitoring of patients at risk remains the best way to prevent and treat thyroid cancer, and current therapies are effective in treating thyroid cancer.

The current gold standard clinically used for diagnosing thyroid cancer is fine needle puncture cytology (FNA) under the guidance of ultrasound, and an article indicates that the diagnosis accuracy is as high as 70% -97%, and most thyroid cancer patients show thyroid nodules found by neck imaging examination. The method comprises the following steps of performing fine needle puncture cytology examination under ultrasonic guidance, namely determining whether a thyroid nodule exists through the ultrasonic examination, then pumping back and forth in the thyroid nodule by using a puncture needle to obtain a small part of diseased tissues, and then obtaining a good cytology smear through a smear so as to judge whether the nodule is good or not. Patients with uncertain FNA biopsy results require thyroidectomy to rule out the possibility of having thyroid cancer, but only 20% of tumors are malignant, meaning that about 80% of patients need to undergo unnecessary surgery, and this method is characterized by difficult sampling, high trauma, etc. Therefore, a rapid, accurate and non-invasive detection means is needed to distinguish good malignant nodules to improve the diagnosis accuracy.

The existing salivary cancer screening is mainly carried out by a Lectin chip, wherein Lectin (Lectin) is a carbohydrate-binding protein which is of non-immune source and has no enzymatic activity, can specifically recognize and bind a specific sugar chain sequence in monosaccharide or glycan with a special structure, and is a biochip prepared by fixing Lectin of different sources on an epoxidation modified chip substrate.

The existing lectin chip cancer identification method mainly comprises the following steps: selecting a control group, comparing the lectin chip result of a clinical sample to be detected with the lectin chip result of a healthy control group to obtain the Fold-change value of each lectin, taking Fold-change >1.5 and Fold-change <0.67 as selection standards, using Fold-change >1.5 as an up-regulated sugar chain and Fold-change <0.67 as a down-regulated sugar chain, screening sugar chain combinations with up-regulation and down-regulation expression, and judging a certain cancer by using the sugar chain combinations. The lectin data applied by the method is not comprehensive, only the lectin data showing the sugar chain structure up-regulation or down-regulation is applied, and the practical application value is lacked under complex situations. Therefore, a papillary thyroid carcinoma recognition system with high accuracy is urgently needed to be developed.

Disclosure of Invention

To overcome the deficiencies of the prior art, the present invention provides a tool and system for testing papillary thyroid cancer that addresses at least one of the aforementioned technical problems.

Specifically, the technical scheme is as follows:

a tool for testing papillary thyroid carcinoma comprising:

a processor;

a storage medium for data interaction with the processor, wherein the storage medium is used for executing the following steps when a program stored in the storage medium is loaded by the processor:

and identifying the lectin chip data of saliva of the patient to be diagnosed by a papillary thyroid cancer identification model, and determining whether the patient to be diagnosed is the papillary thyroid cancer patient.

The tool for testing papillary thyroid carcinoma further comprises:

the collection module collects an external sample and is used for obtaining a pretreated saliva sample:

the pretreatment module is connected with the acquisition module and used for receiving the saliva sample pretreated by the acquisition module to obtain the lectin chip data;

the preprocessing module performs data interaction with the storage medium and stores the lectin chip data in the storage medium.

A system for testing papillary thyroid carcinoma comprising:

the data acquisition module is used for acquiring lectin chip data according to the saliva sample;

the model building module is in data interaction with the data acquisition module and is used for forming classification labels and characteristics by utilizing the lectin chip data so as to build and train a papillary thyroid carcinoma identification model;

the tool for testing papillary thyroid carcinoma of

claim

1 or 2, wherein the data acquisition module is used for performing data interaction with the data acquisition module, and is used for identifying saliva of a patient to be diagnosed by using the trained papillary thyroid carcinoma identification model to determine whether the patient to be diagnosed is a papillary thyroid carcinoma patient.

The model building module comprises:

the data preprocessing unit is in data interaction with the data acquisition module and is used for carrying out normalization processing on the lectin chip data to obtain normalized lectin chip data;

the data classification unit is in data interaction with the data preprocessing unit and is used for randomly sequencing the normalized lectin chip data to obtain the characteristics and the labels of the normalized lectin chip data, and randomly selecting one part of the normalized lectin chip data as a training set of the papillary thyroid cancer identification model and the other part of the normalized lectin chip data as a test set of the papillary thyroid cancer identification model according to the labels;

and the construction unit is connected with the papillary thyroid carcinoma test tool and is used for obtaining the trained papillary thyroid carcinoma identification model and storing the identified papillary thyroid carcinoma identification model in the papillary thyroid carcinoma test tool.

The data preprocessing unit is used for acquiring the data of the data acquisition module and performing the following steps so as to obtain the normalized lectin chip data:

collecting the median fluorescence signal, the median background value, and the standard deviation of the background value of the binding sites of the lectin and the saliva sample;

carrying out screening pretreatment on effective data through the median of the fluorescence signal, the median of the background value and the standard deviation of the background value;

and (3) calculating a median value of three points corresponding to each lectin, eliminating negative quality control and positive quality control, and performing normalization treatment to obtain the lectin chip data.

The "pretreatment of screening effective data by the median of fluorescence signal, the median of background value, and the standard deviation of the background value" includes:

and subtracting the median of the background value from the median of the fluorescence signal, continuously comparing the median with 2 times of the standard deviation of the background value, and taking the data with the standard deviation more than twice of the background value as effective data, otherwise, taking the data as 0.

The construction unit is connected with the data classification unit and used for constructing the papillary thyroid cancer identification model according to the lectin chip data through any one of a K nearest neighbor algorithm, a support vector machine, a multilayer perceptron, logistic regression and random forest by utilizing the features and the labels.

The construction unit receives the tag and the feature for constructing the papillary thyroid cancer identification model from the lectin chip data by:

constructing N SVM (support vector machine) as required, wherein N is a positive integer and is more than or equal to 1;

inputting the characteristics in the data classification unit into the SVM (support vector machine), and obtaining the category of the sample through voting.

The model building module further comprises: a hyper-parameter selection and optimization unit;

the parameter selection and optimization unit is in data interaction with the construction unit for adapting the papillary thyroid carcinoma identification model by:

setting a punishment coefficient of the SVM support vector machine;

when kernel function selection is carried out, a Radial Basis Function (RBF) is adopted as a kernel function of the SVM, and an optimal parameter Gamma under the kernel function is set.

The data acquisition module comprises:

the collecting unit is used for collecting a saliva sample;

the marking unit is connected with the collecting unit and is used for carrying out fluorescence marking on the saliva sample;

the lectin chip unit is connected with the marking unit and is used for placing the saliva sample after the fluorescent marking;

the incubation unit is connected with the lectin chip unit and is used for incubating the lectin chip unit with the fluorescence-labeled saliva sample;

the data acquisition unit is connected with the incubation unit and used for scanning the incubated lectin chip unit and then carrying out image analysis to acquire the lectin chip data;

and the data acquisition unit is connected with the data preprocessing unit and is used for acquiring the normalized lectin chip data.

The invention has at least the following beneficial effects:

the tool for testing papillary thyroid carcinoma can identify saliva of a patient to be diagnosed, and the papillary thyroid carcinoma diagnosis module is used for determining whether the patient to be diagnosed is the papillary thyroid carcinoma patient; the tool for testing thyroid papillary carcinoma is convenient to sample, high in sensitivity, suitable for screening early and middle stage thyroid papillary carcinoma of patients and capable of rapidly identifying whether a subject suffers from the thyroid papillary carcinoma.

According to the system, after the saliva sample is pretreated and fluorescently labeled through the data acquisition module, the lectin chip detection is carried out by using the spotted lectin chip so as to obtain the lectin chip data; identifying saliva of a patient to be diagnosed through a trained papillary thyroid cancer identification model, and determining whether the patient to be diagnosed is a papillary thyroid cancer patient; the system disclosed by the invention comprehensively uses lectin data, combines a machine algorithm, and has the advantages of objectivity and accuracy in detection.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and those skilled in the art can also obtain other related drawings based on the drawings without inventive efforts.

FIG. 1 is a system block diagram of the system of the present invention;

FIG. 2 is a schematic view of a lectin chip;

FIG. 3 is a schematic view of a sample bound to a lectin detection spot;

FIG. 4 is a flow chart of data analysis of lectin chip results;

FIG. 5 is a confusion matrix diagram of the KNN model;

FIG. 6 is a ROC plot of a KNN model;

FIG. 7 is a confusion matrix diagram of an SVM model;

FIG. 8 is a ROC plot of an SVM model;

FIG. 9 is a confusion matrix diagram of the MLP model;

FIG. 10 is a ROC plot of an MLP model;

FIG. 11 is a diagram of a confusion matrix for the LR model;

FIG. 12 is a ROC plot of an LR model;

FIG. 13 is a diagram of a confusion matrix for the RF model;

FIG. 14 is a ROC plot for the RF model;

100, a data acquisition module; 200. a model building module; 300. papillary thyroid carcinoma test tools;

101. a collection unit; 102. a marking unit; 103. a lectin chip unit; 104. an incubation unit; 105. a data acquisition unit;

201. a data preprocessing unit; 202. a data classification unit; 203. a building unit; 204. a parameter selection and optimization unit;

wherein, a in fig. 3 represents a fluorescence signal F532; and B represents a background value B532.

Detailed Description

Those skilled in the art can understand that the modules in the device in the implementation scenario may be distributed in the device in the implementation scenario according to the implementation scenario description, and may also be located in one or more devices different from the present implementation scenario with corresponding changes. The modules of the implementation scenario may be combined into one module, or may be further split into a plurality of sub-modules.

In order to solve the defects in the prior art, the machine learning is introduced to identify papillary thyroid carcinoma in the embodiment, so that the practical application problem of lectin chip data can be perfectly solved.

The specific embodiment is as follows:

the present invention provides an embodiment:

referring to fig. 1, a system for testing papillary thyroid cancer comprises: a data acquisition module 100 for acquiring lectin chip data from a saliva sample, a model construction module 200, and a papillary thyroid carcinoma test tool 300 as described above; wherein, the model construction module 200 performs data interaction with the data acquisition module 100, and is used for forming classification labels and characteristics by using the lectin chip data, so as to construct and train a papillary thyroid carcinoma identification model; the papillary thyroid cancer testing tool 300 as described above, which interacts with the data acquisition module 100 to identify saliva of a patient to be diagnosed by using a trained papillary thyroid cancer identification model, and determines whether the patient to be diagnosed is a papillary thyroid cancer patient; specifically, the data acquisition module 100 is configured to acquire a sample, and the model construction module 200 is configured to perform data processing on the sample acquired by the data acquisition module 100 to obtain a papillary thyroid carcinoma identification model.

Specifically, the model building module 200 includes: the system comprises a data preprocessing unit 201, a data classifying unit 202, a constructing unit 203 and a parameter selecting and optimizing unit 204; the data preprocessing unit 201 performs data interaction with the data acquisition module, and is used for performing normalization processing on the lectin chip data to obtain normalized lectin chip data; the data classification unit 202 performs data interaction with the data preprocessing unit 201, and is configured to randomly sort the normalized lectin chip data, obtain features and labels of the normalized lectin chip data, randomly select a part of the normalized lectin chip data as a training set according to the labels, and select another part of the normalized lectin chip data as a test set; the construction unit 203 is connected to the data classification unit 202, and is configured to construct the papillary thyroid carcinoma identification model according to the features and the labels, and train the papillary thyroid carcinoma identification model by using the training set, so as to obtain the trained papillary thyroid carcinoma identification model; the parameter selection and optimization unit 204 performs data interaction with the construction unit 203, and is configured to successively adjust the parameters and the maximum feature numbers that have the greatest influence in the papillary thyroid cancer identification model, and adjust the papillary thyroid cancer identification model.

The data acquisition module 100 includes: a collection unit 101, a labeling unit 102, a lectin chip unit 103, an incubation unit 104, and a data acquisition unit 105; wherein, the collecting unit 101 is used for collecting saliva samples; the marking unit 102 is connected with the collecting unit 101 and is used for carrying out fluorescence marking on the saliva sample; the lectin chip unit 103 is connected with the marking unit 102 and is used for placing the saliva sample after fluorescent marking; the incubation unit 104 is connected with the lectin chip unit 103 and is used for incubating the lectin chip unit 103 with the fluorescence-labeled saliva sample; the data acquisition unit 105 is connected to the incubation unit 104, and is configured to scan the incubated lectin chip unit 103, and perform image analysis to acquire the lectin chip data.

The lectin chip unit 103 shown in FIG. 2 is prepared as follows:

the lectin chip is selected from: epoxysilane reagent (GPTS), Bradford reagent, protease inhibitor, DMSO, Tween-20, hydroxylamine hydrochloride were all obtained from Sigma, Cy3, Cy5 fluorescent dye from Amersham, Sephadex G-25 column from GE Healthcare, 384 well plate from Genetix, UK, Bovine Serum Albumin (BSA) from Calbiochem, Germany, glass slides from Gold Seal, 0.2 μm filters, 0.45 μm filters from Sartorius, and other commonly used reagents from domestic analytical purity. 37 lectins, see Table 1, were purchased from Vector and Sigma, respectively.

TABLE 1 lectin comparison Table

Apparatus for use, comprising: electric heating air blast drying oven: tianjin Tester Co; an autoclave: TOMY, Japan; ultrafreeze centrifuge 5804R: eppendorf, Germany; micro nucleic acid protein determinator: implen, germany; biochip scanner 4000B: axon corporation, USA; chip sample application instrument: a boso crystal core SmartArrayer48 sample applicator; chip hybridization box HL-2000: UVP, USA.

A process for preparing a lectin chip, comprising:

the untreated slides were washed three times with absolute ethanol for 10min each. After centrifugal drying, the slide is soaked into 250mL of 10% NaOH solution, and the reaction is carried out by shaking on a shaking table and keeping out of the sun overnight. After the reaction, the ultrasonic treatment is carried out for 15min, and then the reaction product is cleaned by ultrapure water for four times, each time lasts for 2min, and cleaned by absolute ethyl alcohol for two times, each time lasts for 2 min. After centrifugal drying, the slide is soaked into 200mL of 10% GPTS solution, and the mixture is shaken on a shaking table and is reacted for 3 hours in a dark place. After the reaction, ultrasonic cleaning is carried out for 15min, and absolute ethyl alcohol is used for cleaning for three times, wherein each time is 10 min. And after centrifugal drying, finishing the epoxidation modification of the chip, and placing the modified glass slide in a drier at 4 ℃ for storage for later use. The prepared lectin chips were designed as spot samples in FIG. 1, each chip was divided into 4 matrices, each matrix was 12 x 10 in size, and each sample spot was repeated three times.

When the collection unit 101 collects a saliva sample, the collection unit comprises:

healthy Volunteers (health Volunteers, HV), without other underlying disease, did not take any medication within a week; patients with well-diagnosed Benign Thyroid Nodule (BTN), and patients with Papillary Thyroid Carcinoma (PTC). Two hours after meal, about 9 to 10 hours, after rinsing three times with normal saline, the whole saliva secreted naturally is collected rapidly. Saliva was collected at least 1mL and immediately placed on ice, protease inhibitor was added at 1 μ L per mL of saliva to prevent protein degradation.

TABLE 2 clinical sample information Table

The working process of the marking unit 102 is as follows: the collected whole saliva was centrifuged at 12,000rpm at 4 ℃ for 10min, and the supernatant was aspirated to discard the precipitate. The supernatant was filtered through a 0.22 μm pore size filter to remove bacteria and other microorganisms. The sample was labeled with Cy3 fluorochrome and free fluorescence was removed by Sephadex G-25 desalting column. The labeled protein is ready for lectin chip incubation.

The working process of the incubation unit 104 is: taking out the prepared lectin chip from the drier at 4 deg.C, and heating; firstly, the chip is washed by PBST and PBS once respectively, each time is 3min, and then centrifugal drying is carried out. The lectin chip was incubated with 600. mu.L of blocking buffer in a chip hybridization cassette and spun at 25 ℃ for 1 h. After the sealing is finished, the chip is washed twice by PBST and PBS respectively, and is dried after 3min each time. The chips after blocking were scanned with a Genepix4000B chip scanner and checked for blocking effect.

Mu.g of the fluorescence-labeled sialoprotein was mixed with the incubation buffer to prepare a 600. mu.L loading system. And uniformly loading on a cover glass, covering with the closed lectin chip, and performing rotary incubation in a chip hybridization instrument at 25 ℃ for 3 hours in a dark place. After incubation, washing the slide twice with PBST and PBS respectively, each time for 5min, and centrifugally drying; during the operation of the data acquisition unit 105, the Genepix4000B chip scanner is used to scan the chip, and the GenePix3.0 software performs circle-point derivation of the GPR file from the chip scanning result chart, and performs analysis according to the data information in the GPR file.

The data information is preprocessed by the data preprocessing unit 201:

as in fig. 3, first the median of the fluorescence signals (F532 mean) of the binding sites (circles) of the collected lectins to the sample, the median (B532 mean) of the background values (the portions of the squares excluding the circles), and the standard deviation (B532 SD) of the background values;

secondly, screening and processing effective data, subtracting the median of a background value from the median of the fluorescence signal, continuously comparing the median with the standard deviation of 2 times of the background value, and considering the data more than two times of the standard deviation of the background value as effective data, otherwise, marking as 0;

then, calculating a median value at every three agglutination points (three points exist in each agglutinin), eliminating negative quality control and positive quality control, and carrying out normalization processing to obtain agglutinin chip data for subsequent analysis; FIG. 4 is a flow chart of data analysis of the lectin chip results;

all cases are randomly ordered by the data classification unit 202, in order to ensure that the proportion of each category in the training set is consistent with that in the test set as much as possible, 70% of data (21 healthy volunteers, 15 benign thyroid nodule patients and 19 thyroid papillary carcinoma patients) in each label (category) are respectively and randomly selected by the invention, 55 cases are spliced and then used as the training set for machine learning, and the rest 30% of data are spliced (24 cases including 9 healthy volunteers, 7 benign thyroid nodule patients and 8 thyroid papillary carcinoma patients) and then used as the test set for machine learning; the three categories may be: healthy, benign thyroid nodule patients, and papillary thyroid carcinoma patients. The specific process is as follows:

a. class names are mapped to numbers:

data["class"]＝data["class"].map({"HV":0,"BTN":1,"PTC":2,})；

b. extracting data of each category:

t0＝data[data['class']＝＝0]

t1＝data[data['class']＝＝1]

t2＝data[data['class']＝＝2]

c. random scrambling of data:

r is 1(r can be any natural number)

t0＝t0.sample(len(t0),random_state＝r)

t1＝t1.sample(len(t1),random_state＝r)

t2＝t2.sample(len(t2),random_state＝r)

d. Data grouping and splicing:

p＝0.70

train_X＝pd.concat_{([t0.iloc[:int(len(t0)*p),2:39],t1.iloc[:int(len(t1)*p),2:39],t2.iloc[:int(len(t2)*p),2:39]],axis＝0)}

train_y＝pd.concat_{([t0.iloc[:int(len(t0)*p),1],t1.iloc[:int(len(t1)*p),1],t2.iloc[:int(len(t2)*p),1]],axis＝0)}

test_X＝pd.concat_{([t0.iloc[int(len(t0)*p):len(t0),2:39],t1.iloc[int(len(t1)*p):len(t1),2:39],t2.iloc[int(len(t2)*p):len} _{(t2),2:39]],axis＝0)}

test_y＝pd.concat_{([t0.iloc[int(len(t0)*p):len(t0),1],t1.iloc[int(len(t1)*p):len(t1),1],t2.iloc[int(len(t2)*p):len(t2),1]],axis} _＝0)

preferably, the construction unit 203 constructs the breast cancer discrimination model by using a Support Vector Machine (SVM); the algorithm idea of the SVM is as follows: the method is essentially a model for realizing two classes, and also finds a linear classifier which can enable the two classes to generate the maximum classification interval on the characteristic space, and can be expanded to the nonlinear situation through a kernel skill.

The implementation steps of the preferred scheme are as follows: inputting case characteristics in a training set, and because the SVM is essentially to realize two classes, an SVM needs to be designed between any two classes of samples. 3-1/2-3 SVM are required to be constructed for realizing the classification of 3 categories, the characteristics of the cases in the test set are input into the trained SVM, the category with the most votes is the category of the unknown sample, and the label prediction and comparison of the test set data are carried out.

Selecting and optimizing hyper-parameters:

c: penalty factor, i.e. tolerance to errors. If C is too large or too small, the generalization ability is poor.

Selection of Kernel function (Kernel): common kernel functions comprise a linear kernel function, a polynomial kernel, a Radial Basis Function (RBF), a Fourier kernel and the like, and a Cross-Validation method is adopted, namely different kernel functions are tried respectively when kernel function selection is carried out, and the kernel function with the minimum induction error is the best kernel function. The best kernel function for this study is RBF. After the RBF function is selected as kernel, the function has a parameter Gamma. The distribution of the data after being mapped to a new feature space is determined, the larger the Gamma is, the fewer the support vectors are, the smaller the Gamma value is, the more the support vectors are, and the larger the RBF width is. The number of support vectors affects the speed of training and prediction. And (3) selecting the optimal parameters of the RBF: the optimal parameters are kernel 1, gamma 11.

The invention discloses an embodiment:

a tool for testing papillary thyroid carcinoma comprising: the device comprises a processor, a storage medium, an acquisition module and a preprocessing module; a storage medium for performing the following steps when the stored program is loaded by the processor: identifying lectin chip data of saliva of a patient to be diagnosed by using a papillary thyroid cancer identification model, and determining whether the patient to be diagnosed is a papillary thyroid cancer patient; for the convenience of use, the collection module collects an external sample, and is used for obtaining a fluorescence-labeled pretreated saliva sample, specifically referring to the "working process of the labeling unit 102" and the "preparation process of the lectin chip unit 103" described in the "system for testing papillary thyroid cancer" example.

The preprocessing module performs data interaction with the acquisition unit and the storage medium, and is used for acquiring lectin chip data, and the specific steps refer to the working processes of the incubation unit 104 and the data preprocessing unit 201 described in the embodiment of the system for testing papillary thyroid carcinoma. Through the lectin chip data, a trained papillary thyroid cancer identification model can be used for identifying a patient to be diagnosed, and whether the patient to be diagnosed is a papillary thyroid cancer patient or not can be determined.

And (3) accuracy verification:

machine learning includes a number of methods, now by: the construction unit 203 compares the modeling results of the logical forest with the modeling results of the logical forest of the embodiment by respectively modeling with a K nearest neighbor algorithm, a support vector machine, a multilayer perceptron and a logistic regression, and checks the accuracy of the embodiment.

The K-Nearest Neighbor (KNN) algorithm is simple, visual, practical and widely applied to the classification problem, and the main idea is to calculate the distance between a point in a known class data set and a current point; sorting according to distance increment; selecting k points closest to the current point; determining the occurrence frequency of the category where the first k points are located; and returning the category with the highest frequency of the first k points as the prediction classification of the current point.

The K nearest neighbor algorithm is realized by the following steps:

constructing a KNN classifier by inputting all the characteristics and labels of cases in a training set; comparing the output prediction result with the label of the test set through the characteristics of the cases in the test set, and measuring the algorithm performance; the most important parameters in KNN comprise a K value, a weight and a distance calculation mode:

k: selecting k points which are closest to the current point, if k is too small and has no anti-interference performance, reducing the deviation bias of the model, increasing the variance and easily overfitting the model; k is too large, which is equivalent to prediction with training examples in a larger neighborhood, and the approximation error of learning increases, and is therefore not representative.

weights: the distance weight is not considered and is considered;

distance measurement mode P: including minkowski distances, euclidean distances, manhattan distances, and the like;

traversal is performed on the values of weights, p and K by using GridSearch, and the result shows that the optimal parameters are weights ═ distance ", p ═ 1 and K ═ 4.

As shown in fig. 5-6 and table 3, the KNN model behaves as follows: the accuracy is as follows: 84.0 percent; the precision ratio is: 82.01 percent; the recall ratio is: 84.24 percent; ROC area below line: 0.88; sensitivity 0.84, specificity: 0.92.

TABLE 3 KNN model data summarization

As shown in fig. 7-8 and table 4, the SVM model behaves as follows: the accuracy is as follows: 92.00 percent; the precision ratio is: 90.48 percent; the recall ratio is: 93.94 percent; ROC area below line: 0.94 of the total weight of the mixture; sensitivity: 0.92, specificity: 0.96.

TABLE 4 SVM model data summarization

Multilayer Perceptron (MLP) algorithm idea: the neural network algorithm is a shallow neural network algorithm and comprises an Input Layer (Input Layer), a Hidden Layer (Hidden Layer) and an Output Layer (Output Layer), wherein each Layer is composed of units, the Input Layer is transmitted by example feature vectors of a training set and transmitted to the next Layer through weights (Weight) of connecting nodes, the Output of the previous Layer is the Input of the next Layer, the number of the Hidden layers is arbitrary, and only one Output Layer and one Input Layer are provided.

The implementation steps are as follows: establishing an MLP model by inputting all characteristics and labels of cases in a training set; and comparing the output prediction result with the label of the test set by testing the characteristics of the cases in the set, and measuring the algorithm performance.

Selecting and optimizing hyper-parameters: there are more than 20 hyper-parameters for MLP, but the most important parameters are the number of hidden layers and the number of neurons in each hidden layer. Therefore, a weight optimization solver (slover), hidden _ layer _ sizes (the number of hidden layers and the number of neurons) is mainly considered in parameter selection. The slover includes 'lbfgs', 'sgd' and 'adam'. 'lbfgs' is a family of quasi-newtonian optimizers that can converge faster and perform better for small data sets. 'sgd' refers to a random gradient descent. 'adam' is a random gradient-based optimizer that works well in terms of training time and validation scores on relatively large data sets.

The selection of the optimal parameters by utilizing GridSearch comprises the selection of an optimizer, the number of hidden layers and the number of neurons: the optimal parameters obtained are slope ═ adam', hidden _ layer _ sizes (7,5,), which include two hidden layers, the first layer is 7 neurons, and the second layer is 5 neurons.

As shown in fig. 9-10 and table 5, the accuracy is: 84.00 percent; the precision ratio is: 82.01%; the recall ratio is: 83.33 percent; ROC area below line: 0.88, sensitivity: 0.84, specificity: 0.92.

TABLE 5 MLP model data summarization

The Logistic Regression (LR) algorithm idea: LR belongs to supervised learning and is a 'classification' algorithm, and the regression essence of LR is that the occurrence probability is divided by the non-occurrence probability and then logarithm is taken.

The implementation steps are as follows: constructing an LR model by inputting all characteristics and labels of cases in a training set; and comparing the output prediction result with the label of the test set by testing the characteristics of the cases in the set, and measuring the algorithm performance.

Selecting and optimizing hyper-parameters:

regularization selection parameter penalty: the alternative values for the penalty parameter are "L1" and "L2", corresponding to the regularization of L1 and regularization of L2, respectively. The optimization algorithm selects a parameter solvent: the solver can only select 'libilinear' when penalty selects L1, and can select libilinear, lbfgs, newton-cg, sag when penalty selects L2; the classification mode selection parameter multi _ class: there are two values of ovr and multinomial that can be selected. In skleran of Python, lrlogistic regression cv uses cross validation to select the regularization coefficient C, so the coefficients of the regularization coefficient C do not need to be optimized; the most preferred parameters are "l2", silver "lbfgs", class _ weight "None, and multi _ class" multinomial ".

As shown in fig. 11-12 and table 6, the LR model behaves as follows: the accuracy is as follows: 76.00 percent; the precision ratio is: 73.54 percent; the recall ratio is: 73.20 percent; ROC area below line: 0.82, sensitivity: 0.76, specificity: 0.88.

TABLE 6 LR model data summarization

Random Forest (RF) algorithm idea: refers to a classifier that trains and predicts samples using multiple decision trees.

The implementation process comprises the following steps: and (3) randomly sampling back m samples from the original training set by using a Bootstrap method, and sampling for n _ tree times. Generating n _ tree training sets; for n _ tree training sets, respectively training n _ tree decision tree models; for a single decision tree model, assuming that the number of training sample features is n, selecting the best feature to split according to information gain/information gain ratio/kini index during each splitting; each tree is known to be split up in such a way that all training examples of the node are known to belong to the same class; forming a random forest by the generated multiple decision trees, and voting according to a classifier of the multiple trees to determine a final classification result; and comparing the output prediction result with the label of the test set by testing the characteristics of the cases in the set, and measuring the algorithm performance.

Selecting and optimizing hyper-parameters: firstly, adjusting a parameter n _ estimators which has the greatest influence on the model, and exploring the n _ estimators by Gridsearch to obtain an optimal value; then, the maximum depth of the tree is adjusted, a smaller max _ depth is searched on the premise of not changing the model effect, the model is simplified, and grid search is carried out to obtain the optimal parameter of the max _ depth; then, the minimum sample number min _ samples _ split required by the internal node subdivision and the minimum sample number min _ samples _ leaf of the leaf node are called together, and the value of the optimal parameter min _ samples _ split and the value of min _ samples _ leaf are obtained; finally, the maximum feature number max _ features is subjected to parameter adjustment to obtain the value of the optimal parameter max _ features.

In this embodiment, the optimum parameter n _ estimators is 34, max _ depth is 3, min _ samples _ split is 2, min _ samples _ leaf is 1, and max _ features is 6.

As shown in fig. 13-14 and table 7, in this example, a papillary thyroid carcinoma identification model was constructed using random forests, and the RF model was represented as follows: the accuracy is as follows: 84.00 percent; the precision ratio is: 82.01%; the recall ratio is: 84.24 percent; ROC area below line: 0.88, sensitivity: 0.84, specificity: 0.92.

TABLE 7 summary of RF model data

The results of the above algorithms were counted as in table 8:

TABLE 8 comparison of the effects of different models

As can be seen from table 8, the performance of the SVM support vector machine is more excellent in identifying papillary thyroid carcinoma patients based on lectin chip data, the accuracy in the test set reaches 92.00%, and 9 of 9 HV, 9 of 7 BTN, and 9 of 9 PTC can be correctly distinguished. In addition, the accuracy of RF, MLP and KNN in the test set reached 84.00%, while the LR model performed poorly, with an accuracy of 76.00% in the test set.

Therefore, the system provided by the invention can improve the accuracy of papillary thyroid carcinoma detection, and has the advantage of convenience in sampling due to the fact that saliva is used for detection.

The above disclosure is only a few specific implementation scenarios of the present invention, however, the present invention is not limited thereto, and any variations that can be made by those skilled in the art are intended to fall within the scope of the present invention. The above-mentioned invention numbers are merely for description and do not represent the merits of the implementation scenarios.

Claims

1. A tool for testing papillary thyroid carcinoma comprising:

a processor;

2. The tool for testing papillary thyroid carcinoma according to claim 1, further comprising:

3. A system for testing papillary thyroid carcinoma comprising:

the tool for testing papillary thyroid carcinoma of claim 1 or 2, wherein the data acquisition module is used for performing data interaction with the data acquisition module, and is used for identifying saliva of a patient to be diagnosed by using the trained papillary thyroid carcinoma identification model to determine whether the patient to be diagnosed is a papillary thyroid carcinoma patient.

4. The system for testing papillary thyroid carcinoma of claim 3, wherein said model building module comprises:

the data classification unit is in data interaction with the data preprocessing unit and is used for randomly sequencing the normalized lectin chip data to obtain the characteristics and the labels of the normalized lectin chip data, randomly selecting a part of the normalized lectin chip data as a training set of the papillary thyroid cancer identification model according to the labels, and using the other part of the normalized lectin chip data as a test set of the papillary thyroid cancer identification model;

5. The system for testing papillary thyroid carcinoma of the claim 4, wherein:

6. The system for testing papillary thyroid carcinoma according to claim 5, wherein the "screening pre-processing of effective data by median of fluorescence signal, median of background value, and standard deviation of the background value" comprises:

7. The system for testing papillary thyroid carcinoma of the claim 4, wherein:

8. The system for testing papillary thyroid carcinoma of claim 7, wherein:

9. The system for testing papillary thyroid carcinoma of claim 4, wherein the model building module further comprises: a hyper-parameter selection and optimization unit;

setting a punishment coefficient of the SVM support vector machine;

10. A system for testing papillary thyroid carcinoma according to claim 3, wherein:

the data acquisition module comprises:

the collecting unit is used for collecting a saliva sample;